Sentence embedding
A whole sentence or passage squeezed into a single vector that captures its overall meaning.
Think of it like
Like a book’s Dewey number — one code that says roughly what it’s about so you can shelve like with like.
Example
A search system embeds every FAQ answer as a vector; a user’s question becomes a vector too, and the nearest ones are the best matches.
How it actually works
Instead of one vector per word, a model pools or specially trains to produce one vector per span. These power semantic search, clustering, and retrieval because closeness in the space means closeness in meaning. Quality hinges on the embedding model — a generic one may miss domain nuance.
For product teams
The workhorse behind semantic search and RAG; picking the right embedding model matters more than people expect.
For engineers
A fixed-length vector for a span, from pooling token states or a purpose-trained encoder, compared via cosine or dot product.
Related
- Word embedding — The single-word ancestor.
- Vector database — Where these vectors get stored and searched.
- Cosine similarity — How they’re compared.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome